Bibliographic citations
Arangoitia, J., (2024). Modelos de regresión paramétricos bivariados para el análisis de supervivencia: una aplicación a tiempos de infección y síntomas [Pontificia Universidad Católica del Perú]. http://hdl.handle.net/20.500.12404/26815
Arangoitia, J., Modelos de regresión paramétricos bivariados para el análisis de supervivencia: una aplicación a tiempos de infección y síntomas []. PE: Pontificia Universidad Católica del Perú; 2024. http://hdl.handle.net/20.500.12404/26815
@mastersthesis{renati/538497,
title = "Modelos de regresión paramétricos bivariados para el análisis de supervivencia: una aplicación a tiempos de infección y síntomas",
author = "Arangoitia Fernández Baca, Jorge Víctor",
publisher = "Pontificia Universidad Católica del Perú",
year = "2024"
}
When studies are carried out on new treatments that can be applied to patients suffering from a certain disease, a fundamental factor to evaluate the effectiveness of such treatment is the determination of whether the patient acquired the disease or not, and if he presented symptoms of that disease, or did not. In other words, it is necessary to know (or estimate) the effect that the application of the new treatment had on the time in which the patient acquired the infection and the time in which he began to present symptoms, variables that make it possible to determine if the treatment was able to prevent the disease, or at least slow its spread, and whether it was able to prevent or mitigate the onset of symptoms. It is important to highlight that the study of the time elapsed until the occurrence of an infection or the appearance of symptoms is a particular case of survival analysis, a branch of statistics whose objective is the study of the time elapsed until the occurrence of a event, as well as the effect of variables characteristic of the individuals to whom the event occurs, for example, in the case of patients, the treatment applied to them (the standard or the new), age, gender, among others. This are known as covariates. Thus, the present work proposes two bivariate parametric models based on distributions and statistical methods used in survival analysis, models that will allow studying the joint behavior of time to infection and time to symptoms, considering the intrinsic relationship between both variables. Then, the estimation method to be used will be the accelerated failure time model, a linear regression model in which it is assumed that the logarithm of the infection time and the logarithm of the symptom time are equal to a linear function of the covariates plus an error multiplied by the scale parameter corresponding to each time. With this in mind, there are two errors (one for the time of infection and the other for the time of symptoms) that correspond to the random component of the regression, a component that will be modeled jointly in the following two ways: Assuming that both errors follow a bivariate extreme value distribution. Assuming a copula model, in which it is assumed that each time presents a Weibull marginal distribution, and the dependency relationship of both times obeys a Gumbel copula. Finally, the previous method can be applied to a specific sample in order to estimate the parameters of the assumed distributions, and in this way determine the effect that each of the covariates has on the times of infection and symptoms. In this particular work, the model will be applied in the couple notification study, carried out by Golden in 2005 and whose objective was to verify if a group of patients presented reinfection and symptoms of a previous disease, as well as the effect of a new therapy on such events.
This item is licensed under a Creative Commons License